[Haskell] ANNOUNCE: FPS - FastPackedStrings 0.2

Bulat Ziganshin bulat.ziganshin at gmail.com
Thu Apr 20 06:40:23 EDT 2006


Hello Simon,

Thursday, April 20, 2006, 1:23:42 PM, you wrote:

>>>it's possible to write lib so it
>>>will use any unboxed array and this will work both with ForeignPtr
>>>and ByteArray# implementations

>    - converting arbitrary Ptrs into PackedStrings without copying,
>      including finalizers if necessary (useful for mmap()).

>    - garbage-collectable PackedStrings can be allocated on the heap, no
>      finalizer necessary

>    - PackedStrings can be passed to foreign functions without copying

> ByteArray# only gives you the second two.  You would need two variants
> of the representation to get all three, and that means extra overhead.

Simon, you skipped the whole idea - the lib should work with any array,
including StorableArray too. it should be just Array->ListLike
interface converter, with additional code specialized for unmovable
arrays. plus UTF-8 support at the top of this:

module ArrayToListLike
class ListLike
  head
  tail
  ...
instance (Array a e) => ListLike (a e)

module UTF8PackedString
import ArrayToListLike
newtype UTF8PackedString = UTF8PackedString (UArray Int Word8)
instance ListLike UTF8PackedString


well, i'm not sure that this can be implemented with classes. but at
least this can be implemented with preprocessor (that is Haskell's
module system ;) just like yourself implemented packed strings for
1/2/4-byte wide representations. here, array type and array elements
type will be parameters:

#define ARRAY UArray
#define ELEMENT Word8
#include "Str.hs"


#define ARRAY StorableArray
#define ELEMENT Word8
#include "Str.hs"


> Optimising the library for GHC 6.4.x isn't a goal (for me, at least).

1. i think that we will have really stable release of 6.6 at
Fall'2007, not earlier

2. ForeignPtr has the following drawbacks even for 6.6:

- need more memory. main goal of using packed string is to reduce
memory requirements so adding 10-15 bytes is bad idea

- can point only to the unmovable memory blocks that may increase
overall memory consumption for some apps. general Array can work with
any memory - movable heap, unmovable heap, or C-allocated

-- 
Best regards,
 Bulat                            mailto:Bulat.Ziganshin at gmail.com



More information about the Libraries mailing list